In Calphad database development, finding the correct setup of the experimental data to inform the optimization process typically is a challenging iterative process of changes of selection and setup of weight factors for given experiments, the setup of appropriate starting values for the database parameters, and subsequent computationally aided iterative minimization of a chosen parameter set. For a scientifically oriented assessor, the expertise and knowledge to acquire with respect to the appropriate tools is significant, and for an optimization to yield a robust and reliable modeling, typically years of experience with general thermodynamic calculations are needed.

Determination of the critical parameter selections and boundaries for the automated optimizations to this day remains a highly manual process in a deceivingly challenging parameter space, as we will demonstrate with the superficially simple Cr-Ni system assessment.

We will showcase the advantages that re-conceptualization of experimental input for optimizations (e.g. a liquidus temperature data point) into a ‘driving-force’ based calculation can have with respect to the stability of the results and the convergence of the optimization process. The consequences that stem from this change with respect to weighting and assessing experimental data will be explored.